Estrogen receptor positive (ER+) breast cancers exhibit highly variable responsiveness to endocrine therapy. Our inability to predict accurately the clinical course of the disease leads to inefficient clinical algorithms associated with high morbidity and health care costs. An analysis of the ER+ breast cancer genome reveals a wide range of somatic mutations and gene copy number changes that drive abnormalities in gene expression and function that underlie the diverse phenotypes observed clinically. To speed the creation of molecular classifications that can be translated into clinically useful prognostic and predictive tests we have executed a series of neoadjuvant endocrine therapy trials. The data from the initial studies designed over a decade ago demonstrated that an analysis of the cellular response to endocrine therapy within the first few weeks to months of treatment can be used to determine the long term prognosis of ER+ breast cancer under treatment with adjuvant endocrine therapy. In particular, measurements of proliferation (Ki67) and ER status of on treatment samples provides prognostic information that is independent of pathological stage and grade and forms the basis for a preoperative endocrine prognostic index (PEPI). In the initial phase of this R01, we extended this pioneering work through an analysis of tumor samples from a Phase 2 trial of preoperative letrozole for clinical stage 2 and 3 ER+ breast cancer (POL study) and identified both pre-treatment and on- treatment gene expression signatures that predict response to aromatase inhibitor (AI) treatment. To validate the signatures discovered in the POL trial and to deepen our biomarker discovery efforts we present a new research plan based on the ACOSOG Z1031 study, a randomized comparison between the three approved AIs in the neoadjuvant setting that is due to complete accrual of 375 patients by mid-2009. In Aim 1 we propose to validate two on-treatment prognostic signatures, the simple PEPI approach based on ER and Ki67 and a 50 gene qRT-PCR intrinsic subtype assay (PAM50) that has already been shown to also predict poor response and relapse when applied to letrozole exposed tumors in the POL study. To address the criticism that a baseline expression signature would still be preferable from a practical standpoint, in Aim 2 we will use mRNA gene expression profiling to further discover and validate pretreatment biomarkers that predict poor responsiveness to AI therapy and elevated relapse risk. In Aims 3 and 4 we will determine the fundamental molecular basis for endocrine therapy resistance by analyzing gene copy aberrations and mutations associated with biological markers of AI resistance. In the last year of the grant we will reanalyze the entire data set when relapse-free survival data becomes available and create a final model that potentially uses all three data types (expression, gene copy and mutation) to optimally predict outcomes. This application will deliver a suite of formalin-fixed tissue-tolerant biomarker approaches that can be further validated in the context of randomized adjuvant therapy trials. Our long term goal is to generate a luminal breast cancer atlas that is annotated for clinical outcomes so that the treatment of ER+ breast cancer can be individually tailored with confidence and the emergence of new targeted treatments can be accelerated.